In Australia there is no single rule that says every business must announce its AI use to every client. But several rules can require it in practice: privacy law, consumer law and professional duties. A new privacy transparency obligation starts on 10 December 2026. The safe position is to tell clients plainly where AI touches their information or their service. You can only do that accurately if you know which AI tools your staff actually use.
That last point is where most businesses get stuck. Clients, tenders and questionnaires ask what tools you use, and the honest answer depends on what happens across the whole team, not on what the policy says.
CalibreforumKnow Where AI Is UsedCalibre Forum’s AI readiness assessment surveys employees and produces a scored report with a 90-day action plan for your business.VisitTL;DR
- No blanket Australian law requires telling clients about AI, but the Privacy Act, consumer law and some professional rules create disclosure duties in specific situations.
- From 10 December 2026, privacy policies of covered organizations must describe certain automated decisions that significantly affect individuals.
- Silence can be misleading under consumer law if a client reasonably expects to be told.
- Say what AI does, what data it sees, where a human checks it and how to reach a person. Keep it short and specific.
- Your answers are only as good as your knowledge of real tool use. Survey staff before you answer clients.
Table of Contents
- What the law actually requires
- A quick test: do you need to tell this client?
- What good disclosure looks like
- How to answer a client AI questionnaire
- The gap: policy versus what staff actually do
- Disclosure by situation
- What the government expects beyond the law
- Our view
- Get the facts before you answer clients
- FAQ
- Sources
What the law actually requires
There are three places where disclosure can become a duty. None is a general “label everything” rule.
Privacy law
The Privacy Act 1988 and its Australian Privacy Principles (APPs) apply to organizations covered by the Act. Businesses with annual turnover of $3 million or less are generally exempt under the small business exemption, though exceptions apply. For covered organizations, the Office of the Australian Information Commissioner (OAIC, the privacy regulator) says privacy policies and notifications should carry clear, transparent information about AI use, and that public-facing tools such as chatbots should be clearly identified as AI.
The OAIC also recommends, as best practice, that organizations do not enter personal information, particularly sensitive information, into publicly available generative AI tools. Under APP 6, personal information can be used or disclosed for the purpose it was collected, or a related purpose the person would reasonably expect. Otherwise you need consent or another exception. If client data goes into an AI tool in a way the client would not expect, you are in consent territory.
The December 2026 transparency obligation
The Privacy and Other Legislation Amendment Act 2024 added a transparency obligation that commences on 10 December 2026. Covered organizations must say in their privacy policies:
- the kinds of personal information used by computer programs;
- the kinds of decisions made solely by those programs;
- the kinds of decisions where the program does something substantially and directly related to making the decision.
It applies where the decision could reasonably be expected to significantly affect a person’s rights or interests. The OAIC’s fact sheet on APP 1.7 to 1.9 explains that the test is about decisions using personal information, and that trade secrets and commercially sensitive information are excluded.
A drafting tool that helps write a newsletter is unlikely to trigger this. A program that screens applicants, scores credit risk or decides who gets a service is a different matter.
Consumer law
The Treasury’s final report on AI and the Australian Consumer Law (October 2025) found existing consumer law fit for purpose and recommended no new AI disclosure rule. It also noted that silence or omission can be misleading where a consumer reasonably expects a fact to be disclosed, and that businesses should make the limitations of AI clear when they use it.
So the test is expectation. If a client would reasonably assume a person wrote and checked their advice, report or reply, and a machine did the work unchecked, saying nothing is the risk.
Professional duties
Some professions add their own layer. The Tax Practitioners Board states that tax practitioners must obtain client permission before disclosing client information to a third party, which may include entering it into AI models and tools. It adds that even locally hosted AI may still require disclosure depending on the circumstances. If you work in law, accounting, health, finance or another regulated field, check your own regulator’s guidance as well. Your professional body is the right place for that question.
A quick test: do you need to tell this client?
Work through these in order for each AI use.
- Does personal or confidential client information go into the tool? If yes, check what the client would reasonably expect, whether your privacy policy says so, and whether consent is needed.
- Does the client talk to the AI directly? Chatbots, voice agents and auto-replies should be identified as AI, with an easy route to a person.
- Does the AI make or shape a decision that significantly affects the client? Then the new APP 1.7 obligation may apply, and a human should be able to review.
- Would the client be surprised or misled by silence? If yes, tell them, whatever the other answers.
- Does a contract or tender require it? Many do, so read the clause.
If all five come back “no”, telling clients is optional. Many businesses still choose to, because the rules are moving and clear disclosure builds trust.
What good disclosure looks like
The OAIC’s position is that disclosure should be meaningful to a reasonable person, not technically exhaustive. Its fact sheet says to avoid overwhelming people with granular detail and to structure information so people can ask for more. Corrs Chambers Westgarth’s analysis adds that generic, high-level statements will not be enough for the APP 1.8 disclosures.
In practice, a good statement covers five things:
- What AI is used for. “We use AI to draft first versions of project summaries” beats “we leverage advanced technologies”.
- What data it sees. Say whether client information is entered, and whether the tool provider may use it for training.
- Who checks the output. Name the role: a consultant reviews every report before it reaches you.
- Where it does not apply. If AI never makes decisions about clients, say so.
- How to reach a person or ask questions. One email address or phone number is enough.
Where to put it
A website AI statement, a line in your privacy policy, a short label on AI-assisted content and a notice in any chatbot cover most small and mid-sized businesses. Add a sentence to engagement letters or proposals where client information is involved. Keep them consistent, because a client who reads the website and the contract will compare them.
A worked example
A 25-person marketing agency uses a generative AI tool to draft social posts, and a transcription tool to summarize client calls. The posts contain no client personal data, so a general website statement is enough. The call summaries contain client staff names and commercial plans, so the agency adds a line to its engagement letter: calls may be transcribed and summarized by an AI tool, a team member checks the summary, and the client can opt out. Two tools, two different disclosure levels, matched to what each does with client information. For recorded calls, see our guide to AI meeting note takers and consent in Australia.
How to answer a client AI questionnaire
Larger clients, government buyers and insurers increasingly send questionnaires asking which AI tools you use, what data goes in and who is accountable. Automating security questionnaires can speed up the formatting, but the facts in the answers have to come from you. Here is the routine.
- Build an inventory. List each tool, what it is used for, which teams use it and what data goes in. A simple table is enough.
- Name an owner. One person accountable for AI use makes the “who is responsible” question easy. A 2026 Australian compliance guide suggests this alongside human oversight for anything that affects a customer.
- Check your answers against reality. Policies list approved tools. Staff often use others. If the questionnaire asks “do staff use unapproved tools?”, you need an honest basis for your answer.
- Describe controls, not intentions. “Staff must not paste client personal information into public AI tools” is a policy. Say whether it is communicated, how, and how you know it is followed.
- Answer at the right level. Give clear, high-level information. Do not volunteer detail you cannot back up, and do not claim certifications or legal compliance you do not hold.
- Keep a dated record. Store the answers so the next questionnaire is consistent and you can show what was true when you answered.
The OAIC notes that software vendors supplying enterprise products used for significant decisions should give their customers clear, high-level information about how the software can be used to make decisions. Ask your vendors for it, and use it in your own privacy policy.
The gap: policy versus what staff actually do
Step three above is where businesses most often get caught. A policy tells you what staff should do. It does not tell you that someone in finance uses a free chatbot to tidy up client spreadsheets, or that a browser extension is summarizing every email.
Our pages on shadow AI in the workplace and AI data leakage cover how this happens. The point for disclosure is simple. If you tell a client “we do not put your data into public AI tools” and a staff member does, the statement is wrong, and under consumer law a wrong statement is a bigger problem than none.

Staff also answer differently when they fear blame. That is why an anonymous survey, where nobody is identified, tends to give a more honest picture than asking in a meeting or auditing individuals. Our guide on running an AI staff survey honestly and anonymously sets out how.
Disclosure by situation
| Situation | Disclose to clients? | What to say |
|---|---|---|
| Chatbot or voice agent answers client queries | Yes, identify it as AI | “You are chatting with an AI assistant. Ask for a person at any time.” |
| Staff paste client personal information into an AI tool | Check privacy policy and expectations; consent may be needed | Which tools, what data, how it is protected |
| AI drafts internal documents, no client data | Usually optional | A general website statement |
| AI screens, scores or decides something significant about a client | Likely required from 10 December 2026 for covered organizations | Kinds of decisions and data in your privacy policy |
| Regulated profession handling client affairs | Check your regulator; permission may be needed | Consent in the engagement terms |
What the government expects beyond the law
The Government’s Guidance for AI Adoption, published 21 October 2025, sets out six practices covering accountability, risk management, data governance, testing, transparency and human oversight. It is voluntary, but enterprise buyers increasingly use it as a reference when they send you questionnaires. That is why the inventory and owner steps above matter even when no law forces disclosure.
For policy wording, see our guides to AI policy in Australia and what applies to your business under AI regulations. For what staff can paste into tools, see the Privacy Act and AI tools.
Our view
Clients rarely object to AI. They object to finding out later, and to answers that turn out to be wrong.
The hard part of disclosure is not the wording. It is knowing what is true. Most owners and managers can tell us what their policy says. Far fewer can say which tools their team actually opens each day, what goes into them and which workflows now depend on them.
That is why we built our assessment around a short, anonymous survey of every staff member. People answer honestly when nobody can be singled out, and the combined answers give management an organization-wide map of tool use and workflows. We do not install software or collect individual data, and we make no compliance or legal claims. The map goes into a scored report with a 90-day action plan, so the next questionnaire or client question is answered from evidence, not from memory.
Start with knowing. Then disclose clearly, in plain words, before anyone has to ask.
Anthony
Get the facts before you answer clients
Our AI readiness assessment gives you a scored report built from your staff’s own anonymous answers, showing current capabilities, risks and a 90-day action plan with recommendations for operational and security improvements. The Report is A$1,990 one-off. The Membership is A$2,388 per year and covers recurring quarterly assessments and reports, which suits businesses whose tool use changes often. You can see what is in the report and how the process works before you start.

If you work in a sector where clients ask detailed questions, we also have pages for medical clinics, aged care and NDIS providers, marketing agencies and IT and managed service providers.
FAQ
Is there an Australian law that says I must tell clients I use AI?
There is no general law requiring it. Treasury’s 2025 review recommended no new AI disclosure rule, but existing laws can require disclosure: privacy law, consumer law where silence would mislead, and professional rules. From 10 December 2026, covered organizations must also describe certain automated decisions in their privacy policies.
Do I need to tell clients when AI writes an email or document?
Usually not by law, if no client personal information is entered and a person reviews the output. The question changes if the client would reasonably expect a person to have written it, or if their confidential information went into the tool. When in doubt, a general statement on your website or in your terms is cheap and clear.
Does the small business exemption mean I can ignore this?
Businesses with annual turnover of $3 million or less are generally exempt from the Privacy Act, but there are exceptions, so check with a privacy lawyer or the OAIC. Even if you are exempt, consumer law still applies, and clients and tenders often ask about AI regardless of your size.
How do I answer a client’s AI questionnaire if I am not sure what my staff use?
Find out first. Build an inventory of tools, uses and data, name an accountable person and check your policy against real behavior. An anonymous staff survey is the quickest way to get an honest picture, and it gives you something accurate to base every answer on.
What should I do about a chatbot on my website?
Identify it as AI at the start of the conversation, give an easy way to reach a person and review transcripts for accuracy and tone. The OAIC says public-facing AI tools such as chatbots should be clearly identified to customers.
Sources
- OAIC: Guidance on privacy and the use of commercially available AI products
- OAIC: Automated Decision-Making Transparency Obligation (APP 1)
- OAIC: APP 1.7-1.9 Transparency Obligation fact sheet
- Treasury: Final report, Review of AI and the Australian Consumer Law
- Tax Practitioners Board: AI and the Code
- Corrs Chambers Westgarth: Transparency in automated decision-making
- Leo Lawyers: AI policies every Australian business needs now
- Legal123: AI Legal Compliance for Australian Businesses (2026 Guide)
- AIRiskAware: AI governance in Australian technology
Recommended
- AI Policy in Australia: What to Include and How to Enforce It
- Privacy Act AI Tools Australia: What Staff Can Paste In
- AI Ethics Policy: A Practical Guide for Australian Businesses
- AI Risk Mitigation: A Practical Guide for Australian Businesses
General information only. Articles on this blog are written with the help of AI writing tools and published by Calibre Forum. They are not legal, insurance, tax or financial advice, so check anything important with a qualified adviser.